Multi-Intensity-Layer PIV application to a Gas Turbine Combustor
Naoki YAMADA, Yuji IKEDA and Tsuyoshi NAKAJIMA
Department of Mechanical Engineering
Kobe University
Rokkodai, Nada, Kobe 657-8501
JAPAN
ABSTRACT
The new technique ‘Multi-Intensity-Layer PIV’ was applied to a gas turbine combustor to
detect the instantaneous droplet distribution and its behavior. Mie scattering theory is applied to the
conventional Particle Image Velocimetry and the technique enables us to detect the instantaneous planar
droplet velocity and size information. This M-PIV technique was evaluated with a generally used
industrial burner under atmospheric pressure and had a good agreement with PDA results. So now on this
study the M-PIV was used to a highly pressurized gas turbine combustor for evaluation comparing with
PDA results. The difference from the previous evaluation with an industrial burner is that the flow field is
pressurized to 0.25 MPa so that many optical error sources exist. We are concentrated into the flow
structure formed by a pilot burner installed in the combustor and as a first step the evaluation is
undertaken in cold flow condition. The purpose of this study is the evaluation of M-PIV within the
application into a gas turbine combustor. Evaluation was done under comparison with size-classified
PDA since this new technique tries to describe droplet behavior and flow structure for each droplet size
classes. The combustion chamber is specially designed for the laser diagnostics in order to undertake the
measurement under realistic operating conditions. The technique is evaluated with the PDA results. The
results indicates that Velocity component comparison indicates the technique has basically good
agreement with PDA results and some discrepancy observed due to dense spray region. In the
downstream region there is the perfect agreement with PDA results because the spray is much more
diluted than the nozzle exit. Additionally, the technique can describe the different flow structure that
depends upon the intensity information. By comparing with PDA results the flow structure of each layer
is able to represent size discriminated flow information. We conclude that the technique is applicable to a
pressurized flow field.
Multi-Intensity-Layer PIV application to a Gas Turbine Combustor
Naoki YAMADA, Yuji IKEDA and Tsuyoshi NAKAJIMA
Department of Mechanical Engineering
Kobe University
Rokkodai, Nada, Kobe 657-8501
JAPAN
ABSTRACT
The new technique ‘Multi-Intensity-Layer PIV’ was applied to a gas turbine combustor to
detect the instantaneous droplet distribution and its behavior. Mie scattering theory is applied to the
conventional Particle Image Velocimetry and the technique enables us to detect the instantaneous planar
droplet velocity and size information. This M-PIV technique was evaluated with a generally used
industrial burner under atmospheric pressure and had a good agreement with PDA results. So now on this
study the M-PIV was used to a highly pressurized gas turbine combustor for evaluation comparing with
PDA results. The difference from the previous evaluation with an industrial burner is that the flow field is
pressurized to 0.25 MPa so that many optical error sources exist. We are concentrated into the flow
structure formed by a pilot burner installed in the combustor and as a first step the evaluation is
undertaken in cold flow condition. The purpose of this study is the evaluation of M-PIV within the
application into a gas turbine combustor. Evaluation was done under comparison with size-classified
PDA since this new technique tries to describe droplet behavior and flow structure for each droplet size
classes. The combustion chamber is specially designed for the laser diagnostics in order to undertake the
measurement under realistic operating conditions. The technique is evaluated with the PDA results. The
results indicates that Velocity component comparison indicates the technique has basically good
agreement with PDA results and some discrepancy observed due to dense spray region. In the
downstream region there is the perfect agreement with PDA results because the spray is much more
diluted than the nozzle exit. Additionally, the technique can describe the different flow structure that
depends upon the intensity information. By comparing with PDA results the flow structure of each layer
is able to represent size discriminated flow information. We conclude that the technique is applicable to a
pressurized flow field.
INTRODUCTION
Droplet dynamics and its interaction to surrounding swirl air flow was investigated using phase
Doppler anemometry (PDA) and particle image velocimetry (PIV). In our previous research we went
through the 4 steps, laser diagnostics error estimations considering with optimizations of optical settings,
combustion chamber optimum designing for optical diagnostics, PDA applications and laser sheet
visualization. In the error estimation process various possible error sources were considered, such as
droplet sticking on the optical windows, alternation of diffractive index due to the ambient pressure
changing, location of the probe volume and slit location changing due to unfixed optical pathways, etc.
Combustion chamber was carefully designed on the next step with considering the area of interest and its
location, required optical pathways, size and location of the optical windows for PDA and PIV
measurements, required thickness of the combustion chamber itself and optical windows that fits for the
ambient pressure up to 5.0 MPa. After these steps PDA was applied to the combustion chamber under
realistic operating conditions. With PDA results time averaged size classified droplet behavior, slip
velocity, scalar distributions such as Reynolds number, drag coefficient, Reynolds stress, vorticity and
dilatation were discussed. Utilizing the high data acquisition feasibility of PDA time scale of the
turbulent and coherent structures was also discussed. And now here in this study the target is focused on
the droplet dynamics under high pressure,
For understanding of the droplet behavior phase Doppler anemometry (PDA/PDPA), with
which temporary averaged droplet size and velocity information is available, was widely used (1). The
technique is good for understanding the time domain information and suitable for acquiring time series
data for statistical analysis at each measurement point so that we are able to know the detailed time
history of velocity and the time scale of the turbulence and coherent structures. For recent years Particle
image velocimetry (2) (3) (4) (PIV) is used to investigate the instantaneous spatial structure since the
technique is based on spatial domain. In recent years spatial resolution is well improved (5) (6) and spatially
detailed information is available. But the problem on PIV for the application to spray field there are no
droplet size information acquired so that we cannot know of which droplet size the acquired velocity is.
For a solution for this problem we proposed “Multi-Intensity-Layer PIV” (7) (8) (9) technique that utilize
Mie scattering theory (10) (11) for PIV and enable us to get droplet size discriminated PIV. This M-PIV
technique was evaluated with a generally used industrial burner under atmospheric pressure and had a
good agreement with PDA results (7) (8) (9). So now on this study the M-PIV was used to a highly
pressurized gas turbine combustor for evaluation comparing with PDA results. The difference from the
previous evaluation with an industrial burner is that the flow field is pressurized to 0.25 MPa so that
many optical error sources (12) (13) are exist (i. e., Thick optical windows on the light pathways, diffractive
index is in vary, etc) We are concentrated into the flow structure formed by a pilot burner installed in the
combustor and as a first step the evaluation is undertaken in cold flow condition. The purpose of this
study is the evaluation of M-PIV within the application into a gas turbine combustor. Evaluation was
done under comparison with size-classified PDA (14) since this new technique tries to describe droplet
behavior and flow structure for each droplet size classes.
EXPERIMENTAL DESCRIPTION
Figure 1 shows a schematic drawing of the test nozzle assembly. High-pressure air passed
through axial swirler vanes surrounding the nozzle, and was then ejected at the swirler exit, which had an
inner diameter of 81 mm. At the tip of the nozzle body, which had an outer diameter of 54 mm, a
pressure-swirl atomizer (DELAVAN Inc.) was mounted on the central axis of the nozzle. The flow rate
of the atomizer was 35 GPH (gallons per hour), and a spray angle of 80° was chosen to create a hollowcone spray. The test nozzle assembly, which had a vertical traverse system, was mounted in a pressure
vessel of 230 mm in inner diameter and 1,300 mm in vertical length. It could operate at pressures up to
5.0 MPa. The combustion chamber was designed with purged quartz windows, through which laser
diagnostics and chemiluminescence measurements of combusting spray could be made from outside the
combustion chamber. This study was done under non-combustion condition so that water was injected
instead of fuel (light oil). Water was pressurized by a fuel pump and sprayed into the vessel. The pressure
vessel was carefully designed, and much attention was paid to the size and location of the optical
windows used to measure phase Doppler anemometry (PDA) and laser simultaneously.
The area of interest extended about 80 mm downstream from the nozzle exit. To measure droplet
size and velocity, an Ar-ion laser and a fiber PDA system (DANTEC) were used. The focal lengths of the
transmitting and receiving optics were both 600mm, both the transmitting and receiving optics were
mounted on a one-dimension horizontal traverse table. The light source for PIV is Nd:YAG laser
(400mJ/pulse) and images are captured by a cross-correlation camera (Kodak ES1.0, 1008H x 1018W
pixels). The experiments were conducted under a fixed air temperature of 20°C, an airflow rate of 0.172
m3/s. and fuel (water) flow rate was 104 kg/h. The ambient air pressure was 0.25 MPa.
Optical
window
Swirler
Φ
Nozzle
Φ
Φ
Fig.1 Experimental apparatus
CONCEPT OF MULTI-INTENSITY-LAYER PIV
The method is a simple application of Mie theory (10), Mie scattering intensity from droplet is
proportional to diameter squared in a certain receiving angle. Generally Mie scattering intensity of 10 to
100µm droplets is believed to be proportional at 30o receiving angle. That means the intensity
information on CCD captured image includes the droplet diameter information so that we proposed to
utilize its intensity information to convert into droplet size information on PIV data analysis.
Fundamental concept of this Multi-Intensity-Layer PIV is shown in Figs. 2 and 3. The light scattered
from the droplets was captured and digitized in 8bit, 0 to 255 grayscale. A couple of successive 8bit
source images, that were illuminated with slight time difference (20µs) each other, were distributed into
three different images depending on intensity information of each pixels. We named these distributed
images as ‘layer’ and then the source image is distributed into three layers. Layer distribution criterion of
8bit image is the importance of this technique and should carefully be done. This M-PIV is to measure
the typical spray behaviors (i.e., following/penetration in a large scale turbulent structure), not trying to
tell detailed droplet diameter from the source images. This technique will enable us to have the
instantaneous spatial spray structure for each droplet size groups in 2-dimensional plane.
Based on the PDA results in the flow field droplet diameter distribution is from nearly 0µm up to
80µm so that we assume that the intensity from nearly 0µm droplets equals 0 in 8bit grayscale and 80µm
droplets to 255. Our previous research tells that less than 30µm droplet follows the large scale turbulent
structure while more than 50µm droplet penetrates, and 30 to 50µm droplets are intermittent. So we tried
to detect the droplet behavior of the size of 0 to 30µm, 30 to 50µm and more than 50µm.
Averaged image
x [mm]
0
Instantaneous image
40
80
0
r [mm]
70
0
r [mm]
70
Fig.2 Averaged and instantaneous images of spray flow field
We define that the intensity from nearly 0µm droplet is 14 since 0 to 13 is the dark current noise on CCD
array, and the intensity of 80µm is equivalent to 255. Based upon Mie theory, 02 equals to 14 of 255
intensity and 802 equals to 255 on 255 grayscale. The intensity of 30µm and 50µm were calculated by
simple linear interpolation. The intensity of source images were finally distributed into three layers,
layer1 (14-47 on 255), layer2 (48-73 on 256) and layer3 (74-255 on 255) based on Mie theory. The
concept of layer distribution is shown in Fig. 3.
RESULTS AND DISCUSSION
Based on the above-mentioned concept, source images used for conventional PIV were
distinguished into three images (layers) and cross-correlation was executed between temporally
successive two images (layers). At each layer conditions 160 vector maps were averaged for statistical
evaluation. Figure 4 shows the vector maps of normal PIV and M-PIV. On the vector map of normal PIV
there are two strong velocity regions, one is from the droplets injected from the nozzle, that consist of
hollow cone structure, and the other is from the droplets that follows the strong air flow from the swirler
vane. Between these strong velocity regions there is a stagnation region around x=0mm, r=20 mm area.
On the other hand, there is quite large recirculation region observed around the central axis. The region is
located within the hollow cone structure so that the pressure is lower enough than outside the cone. That
is the reason there is such large recirculation region formed in this region. On the figure layer 1 to 3
vector maps are also shown. Fundamental flow structure, that is above mentioned, is the same as the
vectors of normal PIV and little difference can be observed only from these vector maps of each intensity
layers.
Source image
(8bit 255 grayscale)
Layer distribution
0 14
47
74
Layer 1
255
Layer 2
Noise
Layer 3
SIgnal
Fig.3 Concept of Multi-Intensity-Layer PIV
Normal PIV
x [mm]
0
20
40
0
M-PIV Layer 1
20
40 r [mm]
M-PIV Layer 2
M-PIV Layer 3
0
0
x [mm]
0
20
40
0
20
40
20
r [mm]
40
20
40
Fig.4 Vector maps by normal PIV and Multi-Intensity-Layer PIV
Axial velocity
50
Radial velocity
x=7.5mm
40
x=7.5mm
30
20
10
velocity (m/s)
0
50
x=15mm
40
x=15mm
30
20
PDA
dp<5
5<dp<10
10<dp<20
20<dp<30
30<dp<50
50<dp<70
70<dp<100
100<dp
M-PIV
10
layer1
0
layer2
50
x=35mm
40
x=35mm
layer3
30
20
10
0
-10
0
10
20
30
r (mm)
40 0
10
20
30
40
r (mm)
Fig.5 Multi-Intensity-Layer PIV evaluation with size-classified PDA
For comparison and evaluation of each layer vectors with PDA results velocity components of
each layer are shown on Fig. 5 together with size-classified PDA results. On the figure axial and radial
velocity components of axial distance x=7.5, 15 and 20 mm level are shown. At x=7.5 mm level the
vectors from higher intensity have the slower velocity in both axial and radial component, as the fact is
the same with the PDA results. But on the level M-PIV velocity detected is stronger than that of PDA
around r=10 to 20 mm. The possible reason is the dense droplet density. The region is just around the
nozzle exit so that droplets are so dense enough to form high intensity. The intensity distributions on
CCD images include both droplet size and density information so that the application of the technique to
dense region sometimes causes such kind of data variations. On x=15 mm level larger droplets have
larger velocity on axial component and larger droplet have the smaller velocity on radial component
based on PDA results. The same tendency was observed by M-PIV so that the layer1, lower intensity,
represents smaller droplets while layer3, higher intensity, represents larger droplets. The velocity
components of axial and radial has basically good agreement but some discrepancies were observed also
around r=10 mm region. The above mentioning droplet dense effect still remains on this level. At x=35
mm level there are fairly good agreements are shown both on axial and radial components. On this level
M-PIV could perfectly detect the droplet velocity profiles.
For the other way for describing the spatial flow structure, vorticity distributions of each layer
are shown on Fig. 6. On the figure the vorticity from normal PIV is also attached. On the vorticity of
normal PIV there are strong vorticity region on the above mentioned stagnation region and around the
exit of swirler vane. The strong vorticity regions are located on this stagnation region and the ambient air
supplied region. On the latter region there are supposed to be exist the strong shear between the air and
droplets. By M-PIV there are different vorticity that could be detected and proofed that the technique is
able to demonstrate the different flow structure depend on droplet size. Vorticity distribution of layer3 is
always lower than that of layer1 and the reason is the larger size droplets, that are normally penetrating
the large scale vortex structures, were detected on layer3. There are little stagnation region and shear
region detected on layer3 while strong ones were detected on layer1 and 2. The important thing to
describe is by distributing the source images based on the intensity information different flow structures
were observed by using this M-PIV technique even in a pressurized flow condition.
Normal PIV
Vorticity
2000
1600
1200
800
400
0
-400
-800
-1200
-1600
-2000
x [mm]
0
20
40
0
M-PIV Layer 1
20
40 r [mm]
M-PIV Layer 2
M-PIV Layer 3
0
0
x [mm]
0
20
40
0
20
40
20
r [mm]
40
20
40
Fig. 6 Vorticity distributions by normal PIV and Multi-Intensity-Layer PIV
CONCLUSIONS
Novel Multi-Intensity-Layer PIV technique is applied to a combustor in an industrial gas turbine
combustor. The combustion chamber is specially designed for the laser diagnostics in order to undertake
the measurement under realistic operating conditions. The technique is evaluated with the PDA results.
Conclusions are as follows:
-
Vector maps of each layer have few difference of each layer and that indicates there are not so big
differences of fundamental flow structures.
-
Velocity component comparison indicates the technique has basically good agreement with PDA
results and some discrepancy observed due to dense spray region. In the downstream region there is
the perfect agreement with PDA results because the spray is much more diluted than the nozzle exit.
-
The technique can describe the different flow structure that depends upon the intensity information.
By comparing with PDA results the flow structure of each layer is able to represent size
discriminated flow information.
-
The technique is applicable to a pressurized flow field.
REFERENCES
(1) Taylor, A M. K. P. 1993, Instrumentation for Flows with Combustion, Academic press.
(2) Raffel, M. et al, 1998, Particle Image Velocimetry, A Practical Guide, Springer-Verlag.
(3) Grant, I., 1994, Particle Image Velocimetry, SPIE Milestone Series Volume MS 99
(4) Measurement Science and Technology, 1997, Vol. 8, No. 12.
(5) Keane, R. D., et al, 1996, Super-resolution Particle Imaging Velocimetry, Measurement Science and
Technology, Volume 6, pp.754-768.
(6) Hart, D. P., 1998, Super Resolution PIV by Recursive Local-Correlation, VSJ-SPIE98, S3-4-5.
(7) Yamada, N., et al, 1998, Multi-Intensity-Layer PIV for Spray Measurement, 9th International
Symposium on Applications of Laser Techniques to Fluid Mechanics, pp.12.6.1-12.6.8.
(8) Ikeda, Y., et al, 2000, T., Multi-Intensity-Layer PIV for Spray Measurement, Measurement Science
and Technology, Volume 11, Issue 6, pp.617-626.
(9) Yamada, N., et al, 2000, Multi-Intensity-Layer PIV application to a practical burner, 10th
International symposium on applications of Laser Techniques to Fluid Mechanics, 30.4.
(10) Van de Hulst, H. C., 1981, Light Scattering by Small Particles, Dover Publications, Inc.
(11) Domann, R. and Hardalupas, Y., 2000, Evaluation of the Planar Droplet Sizing (PDS) Technique,
Eighth International Conference on Liquid Atomization and Spray Systems (ICLASS 2000).
(12) Ikeda, Y. et al, 1997, Measurement Uncertainties of Phase Doppler Technique Due to Effects of Slit
Location, Control Volume Size and Flame Front Presence (in Application for Combusting Spray),
Application of Laser Anemometry to Fluid Mechanics, Springer-Verlag, pp.165-179.
(13) Ikeda, Y. et al, 1996, Measurement Uncertainties of Phase Doppler Technique Due to Effects of Slit
Location, Control Volume Size and Flame Front Presence (in Application for Combusting Spray), 8th
International Symposium on Applications of Laser Techniques to Fluid Mechanics, Vol.1, pp.2.6.12.6.8.
(14) Ikeda, Y., 1997, Size-classified Droplet Dynamics and its Slip Velocity Variation of Air-assist
Injector Spray, SAE International Congress & Exposition, SAE Paper No.970632.